Skip to main content

​
One-shot setup

The fastest way to get protected: 
rampart quickstart
This auto-detects your agent (Claude Code, Cline, or OpenClaw), installs the background service, configures hooks, and runs a health check. Done.
If you have multiple agents installed, rampart quickstart will set up all detected agents.

​
Manual setup

If you prefer to set up step-by-step, or need to set up a specific agent:
1

Install the background service

This installs rampart serve as a boot service (systemd on Linux, launchd on macOS) and saves a bearer token to ~/.rampart/token:
rampart serve install
The service runs on localhost:9090 by default.
2

Wire up your agent

Choose your agent:
rampart setup claude-code
For agents without native hook support, rampart wrap sets $SHELL to a policy-checking shim. Works with Aider, OpenCode, Continue, and any agent that reads the $SHELL environment variable.
3

Verify everything works

Run the health check:
rampart doctor
You should see:
  • ✓ Background service is running
  • ✓ Token is valid
  • ✓ Policies loaded (3+ rules)
  • ✓ Agent hooks configured

​
Test your first policy

Try running a dangerous command through the policy engine: 
rampart test "rm -rf /"
You should see: 
🔴 deny — Destructive command blocked [block-destructive]
Now try a safe command: 
rampart test "git status"

✅ allow [default]
rampart test evaluates commands against your policies without actually executing them. Perfect for testing policy changes.

​
Use your agent normally

Once Rampart is set up, just use your agent as usual. Every command goes through the policy engine first. For Claude Code: 
claude
For wrapped agents: 
rampart wrap -- aider

​
Watch what’s happening

See real-time policy decisions: 
rampart watch
This opens a live TUI dashboard showing:
  • ✅ Allowed commands
  • 🔴 Blocked commands
  • 🟡 Watched commands (logged for review)
  • 👤 Commands waiting for approval
Press q to exit the dashboard. The background service keeps running.

​
Built-in policy profiles

Rampart ships with four policy profiles:
ProfileDefault actionUse case
standardallowBlock dangerous, watch suspicious, allow the rest
ciallowStrict mode for headless/CI — all approvals become denies
paranoiddenyExplicit allowlist for everything
yoloallowLog-only, no blocking
Switch profiles with: 
rampart init --profile paranoid
The paranoid profile denies everything by default. You’ll need to explicitly allow every command your agent runs. Great for high-security environments, painful for day-to-day development.

​
Customize your policy

When a command is blocked, Rampart tells you how to allow it: 
# Command blocked: "npm install lodash"
# 💡 To allow this: rampart allow "npm install *"
Just run the suggested command: 
rampart allow "npm install *"
# ✓ Rule added — policy reloaded (12 rules active)
Common customizations: 
rampart allow "go test ./..."
rampart allow "docker build *"
Custom rules are stored in ~/.rampart/policies/custom.yaml and won’t be overwritten by rampart upgrade.

​
Next steps

Writing policies

Learn YAML policy syntax, glob patterns, and approval flows

Audit trail

Explore hash-chained audit logs and SIEM integration

Project policies

Commit project-specific rules to your repo for team-shared security

Dashboard

Use the web dashboard to approve commands and test policies

Build docs developers (and LLMs) love

Get started for freeTalk to us